應用資料包絡分析法、蜂群演算法與類神經網路以建構股票投資組合最佳策略

本研究以資料包絡法、蜂群演算法及類神經網路等方法建置一系列之投資組合最佳化程序，以風險最低獲利為目標的前題下進行投資活動。首先，運用資料包絡法作為選股之依據，再以蜂群演算法求出投資模型之股票投資權重，利用類神經網路預測股票並擇時作最佳之賣入賣出交易策略，最後的成果再與股票大盤投資報酬率或銀行定存利率作比較，來驗證建置之整合性投資策略之可行性與有效性。以台灣股票市場電腦及週邊產業類股作為實證標的，實證研究資料期間介於2009年到2011年，每半年為一個期間共建置四個個案為例之投資組合最佳化程序。本驗證結果得出四個模型的建構之實證投資報酬率均優於類股投資報酬率，代表本研究所作的投資最佳化組合之建置均優於大盤類股之投資報酬率，但於類股投資報酬率為負報酬率的情況下，雖然整體效益依然有獲利之效益，但若此時不作投資行為，將現金定存於銀行更優於股票投資行為。

關鍵字

資料包絡法；蜂群演算法；類神經網路

並列摘要

This paper uses methods including the data envelope analysis, artificial bee colony and artificial bee colony to build a series of investment portfolio optimization programs to carry out investment activities on the premise of making profits with minimum risk. First, this paper employs the data envelope analysis as the basis for stock selection and the artificial bee colony to obtain the stock investment weights of the investment model before using the artificial bee colony for prediction of the stock and selection of the optimal timing of buy and sell transaction strategies. The final results are compared with the stock market investment return rate or the bank fixed deposit rate to verify the feasibility and effectiveness of the integrated investment strategies. With the stocks in the computer and peripheral industries in Taiwan's stock market as the empirical subjects, in research data period from 2009 to 2011, this paper establishes the investment portfolio optimization programs of four cases at interval of half a year. According to the verification results, the empirical investment return rate in case of the four models is better than the investment return rate of stocks by sector, indicating that the investment return rate of the investment optimization portfolios proposed is better than the market. However, in case of stock sector investment return rate is negative, although the overall performance is profitable, the deposit of cash in the bank will be better than investment in stocks.